[llm_nerd]

Just a few days ago, on Friday, my 15 year old son had his Claude account suspended with a demand for ID to prove he is 18 or older. He had his own Claude Max subscription (he out-earns me fairly frequently in his circle of gaming programmers), and was unaware Anthropic had a must-be-18 rule, as was I. Their email said "Our team found signals that your account was used by a child. This breaks our rules, so we paused your access to Claude." So I guess if you ever ask a question that seems to originate from a teen or less, expect to hit an ID gate.

So now he's a Codex user. OpenAI and Google both have a minimum age of 13.

EDIT: I should note that Anthropic gave him a refund for the whole month that was underway, despite him being nearing the end of it. So good on them.

[nozzlegear]

> he out-earns me fairly frequently in his circle of gaming programmers

Can you expand on this? Your teenage son makes more money than you do professionally, by vibe coding video games?

[llm_nerd]

Who said anything about "vibe coding"? Using coding tools like Claude Code as just another tool in the belt is something the overwhelming bulk of professional devs do now (and given that my son managed to find a number of clients paying for his work, he qualifies as professional). Pejorative "vibe coding" nonsense doesn't change this.

[nozzlegear]

Call it whatever you want then, I'm still interested in the question at hand – your son makes more money than you do professionally, using Claude Code to make something video game related?

[paradox242]

Almost certainly referring to cheating software.

[llm_nerd]

He makes solutions for people and they pay him money for doing so. I mean...pretty much exactly how we all operate? He's excellent at networking and has built an enormous connection tree.

[nozzlegear]

That's interesting, good for him! When I was a teenager I spent all my time playing Halo 2, I didn't write my first line of code until my early 20s.

[jazz9k]

This sounds so sketchy and vague. Why do you need to keep defending it, if it's complely legit?

[llm_nerd]

"Defending it"? Sketchy and vague is similarly hilariously pathetic language, and you sound like an absolute creep.

I have zero obligation to detail the work my minor son does to random weird foot-stomping, entitled creeps on HN, and these bizarrely insecure demands by professional failures is...telling.

In this case I assume you're keying off of the other clown who, based upon absolutely nothing (but apparently their own professional failure), is certain it must be "cheating software".

How pathetic. If this is your lot in life, Jesus Christ find a different career. Maybe the trades or something.

[alistairmayo]

Yeah. I do not get the 18-years-old age gate. It's not like they're protecting anyone. AI is available so freely now anyone who wants it can get it.

Anthropic made the best models by hiring non-technical folks like philosophers to build the best training sets and evaluations. Now, it seems like their philosophers are telling people how they can and can't use their model.

[basisword]

Liability. OpenAI have had several court cases now I believe where children killed themselves after interacting with ChatGPT. Less liability if the user is an adult.

[progbits]

> Anthropic gave him a refund for the whole month that was underway, despite him being nearing the end of it

I sense an opportunity for free tokens.

Ideas for prompts that reliably trigger the age check?

[trollbridge]

... so let me understand this.

It is frequently said that programming directly is obsolete, and the skill you must have now is knowing how to operate agentic AIs.

Yet you aren't allowed to do this until you're 18.

So, developing software is now 18+ only?

[Kim_Bruning]

Qwen3 runs locally on reasonable hardware, and is comparable to a mid-2025 Claude Sonnet (albeit possibly rather slower) .

Local models are chasing the online frontier models pretty hard.

So worst case, that's the fallback (FWIW, YMMV)

edit: Qwen-3.5 MoE (and other local MoE models like it)

[HWR_14]

Whats "reasonable hardware"?

[Someone1234]

People have tried to run Qwen3-235B-A22B-Thinking-2507 on 4x $600 used, Nvidia 3090s with 24 GB of VRAM each (96 GB total), and while it runs, it is too slow for production grade (<8 tokens/second). So we're already at $2400 before you've purchased system memory and CPU; and it is too slow for a "Sonnet equivalent" setup yet...

You can quantize it of course, but if the idea is "as close to Sonnet as possible," then while quantized models are objectively more efficient they are sacrificing precision for it.

So next step is to up that speed, so we're at 4x $1300, Nvidia 5090s with 32 GB of VRAM each (128 GB), or $5,200 before RAM/CPU/etc. All of this additional cost to increase your tokens/second without lobotomizing the model. This still may not be enough.

I guess my point is: You see this conversation a LOT online. "Qwen3 can be near Sonnet!" but then when asked how, instead of giving you an answer for the true "near Sonnet" model per benchmarks, they suddenly start talking about a substantially inferior Qwen3 model that is cheap to run at home (e.g. 27B/30B quantized down to Q4/Q5).

The local models absolutely DO exist that are "near Sonnet." The hardware to actually run them is the bottleneck, and it is a HUGE financial/practical bottleneck. If you had a $10K all-in budget, it isn't actually insane for this class of model, and the sky really is the limit (again to reduce quantization and or increase tokens/second).

PS - And electricity costs are non-trivial for 4x 3090s or 4x 5090s.

[Kim_Bruning]

I may have genuinely new data for you.

Qwen3.5-35B-A3B is reported to perform slightly better than the model you mentioned.

It runs fine but non-optimal on a single 3090 with even 131072 tokens of context , and due to the hybrid attention architecture, the memory usage and compute scale rather less drastically than ctx^2. I've had friends with smaller cards still getting work out of it. Generation is at around 20 tokens/sec on that 3090 (without doing anything special yet) . You'll need enough DRAM to hold the bits of the model that don't fit. Nothing to write home about, but genuinely usable in a pinch or for tasks that don't need immediate interactivity.

It's the first local model that passes my personal kimbench usability benchmark at least. Just be aware that it is extremely verbose in thinking mode. Seems to be a qwen thing.

(edit: On rechecking my numbers; I now realize I can possibly optimize this a lot better)

[Someone1234]

With respect, this isn't "new data" it is an anecdote. And it kind of represents exactly the problem I was talking about above:

- Qwen is near Sonnet 4.5!

- How do I run that?

- [Starts talking about something inferior that isn't near Sonnet 4.5].

It is this strange bait/switch discussion that happens over and over. Least of all because Sonnet has a 200K context window, and most of these ancdotes aren't for anywhere near that context size.

[yencabulator]

One sure would expect Qwen3.5-35B-A3B to "perform slightly better" than Qwen3-235B-A22B!

[zozbot234]

> The hardware to actually run them is the bottleneck, and it is a HUGE financial/practical bottleneck.

That's unsurprising, seeing as inference for agentic coding is extremely context- and token-intensive compared to general chat. Especially if you want it to be fast enough for a real-time response, as opposed to just running coding tasks overnight in a batch and checking the results as they arrive. Maybe we should go back to viewing "coding" as a batch task, where you submit a "job" to be queued for the big iron and wait for the results.

[Borealid]

A machine with 128GB of unified system RAM will run reasonable-fidelity quantizations (4-bit or more).

If you ever want to answer this type of question yourself, you can look at the size of the model files. Loading a model usually uses an amount of RAM around the size it occupies on disk, plus a few gigabytes for the context window.

Qwen3.5-122B-A10B is 120GB. Quantized to 4 bits it is ~70GB. You can run a 70GB model in 80GB of VRAM or 128GB of unified normal RAM.

Systems with that capability cost a small number of thousand USD to purchase new.

If you are willing to sacrifice some performance, you can take advantage of the model being a mixture-of-experts and use disk space to get by with less RAM/VRAM, but inference speed will suffer.

[fy20]

If you want something off the shelf get a MacBook Pro M5 (base "Pro" CPU) with 48GB RAM:

Gemma 4 31B Q6: 9tok/s, I'd say it is smarter than GPT-4o, but yeah it's slow. Good for coding.

Gemma 4 26B A4B Q4: 50tok/s. Feels faster than ChatGPT 5.4, but not as smart (as it reasons less). Good for general chatting and research.

[fortyseven]

Give Gemma4 a look, too. I've had terrific results with that and OpenCode locally.

[square_usual]

> It is frequently said that programming directly is obsolete

Who says this?

[sn0wleppard]

The CEO of the company in question with the age limit, for one

[paradox242]

I would hope most people can recognize that someone trying to sell you something might be among the least trustworthy sources about that thing.

[bryancoxwell]

I mean, you can disagree with the sentiment (I certainly do), but there are still an awful lot of people saying it.

[nozzlegear]

I thought it was true too, for a couple of months. Then the honeymoon phase ended and now I only use Claude to write commit message drafts (which I rewrite myself) and review PRs.

[guzfip]

Yes, today’s kids should instead learn to be influencers.

This is genuine advice I’ve seen from high profile business types. We’re fucked in the sense our children will be made to be attention whores online.

[llm_nerd]

It seems out of step and foolish, and the cynic in me says that Anthropic has a side hustle of identity harvesting and is looking for justifications, but on the flip side, there is a real risk of pearl clutching if a child ever uses AI, and maybe Anthropic just wants to steer clear of all of that. Though simply putting it in the ToS should be sufficient legal shielding, and the idea that they're chat harvesting to age fingerprint conversations seems dubious.